Can the first stage of a Falcon 9 land itself back on Earth?
That question is worth the difference between $62M and $165M per launch.
Predict whether the Falcon 9's first stage booster will successfully land after a launch — a critical input for estimating the true cost of each mission. The project covers the full Data Science lifecycle: raw data collection, wrangling, SQL + visual EDA, interactive maps, a live dashboard, and a four-model machine learning comparison.
01_dataCollection/
| Notebook | Description |
|---|---|
01_APICollection.ipynb |
Pulls Falcon 9 launch records from the official SpaceX REST API → dataset_part_1.csv |
02_WebScraping.ipynb |
Scrapes Falcon 9 & Falcon Heavy tables from Wikipedia using BeautifulSoup → wiki_launches.csv |
02_dataWrangling/
| Notebook | Description |
|---|---|
03_DataWrangling.ipynb |
Handles nulls, encodes categoricals, and engineers the target variable Class (1 = successful landing, 0 = failure) → dataset_part_2.csv |
03_EDA/
SQL Analysis
04_SQL.ipynb— Success patterns, average payload mass, mission counts per launch site
Visual Analysis (Seaborn / Matplotlib)
| Notebook | Analysis |
|---|---|
| 05_01 | Flight Number vs. Launch Site |
| 05_02 | Payload Mass vs. Launch Site |
| 05_03 | Success Rate by Orbit Type |
| 05_04 | Flight Number vs. Orbit Type |
| 05_05 | Payload Mass vs. Orbit Type |
| 05_06 | Yearly Launch Success Trend |
| 05_07–08 | Launch Site Name Exploration |
| 05_09–15 | Payload & Mission Outcome Metrics |
| 05_16–18 | Pie Charts & Success Scatter by Site |
04_mapsDashboards/
| File | Description |
|---|---|
06_InteractiveMapsFolium.ipynb |
Marker clusters 🟢 success / 🔴 failure, coastline distance via Haversine formula |
07_LaunchSiteDashApp.py |
Plotly Dash App — filter by launch site, dynamic performance charts |
05_machineLearning/
| Notebook | Description |
|---|---|
08_MLPrediction.ipynb |
StandardScaler normalization, 80/20 split, GridSearchCV hyperparameter tuning |
09_MLComparison.ipynb |
Accuracy comparison across all four algorithms |
10_confusionMatrix.ipynb |
Confusion matrices and final evaluation metrics |
| Model | Accuracy |
|---|---|
| Logistic Regression | ~83% |
| Support Vector Machine | ~83% |
| Decision Tree | ~89% ✨ |
| K-Nearest Neighbors | ~83% |
Decision Tree achieved the highest accuracy after hyperparameter optimization with
GridSearchCV.
DS_Capstone_Coursera_IBM/
│
├── 01_dataCollection/
│ ├── 01_APICollection.ipynb ← SpaceX REST API
│ └── 02_WebScraping.ipynb ← Wikipedia + BeautifulSoup
│
├── 02_dataWrangling/
│ └── 03_DataWrangling.ipynb ← cleaning + Class variable
│
├── 03_EDA/
│ ├── 04_SQL.ipynb ← SQL queries
│ └── 05_01…18_*.ipynb ← 18 visual analyses
│
├── 04_mapsDashboards/
│ ├── 06_InteractiveMapsFolium.ipynb
│ └── 07_LaunchSiteDashApp.py ← Dash web app
│
├── 05_machineLearning/
│ ├── 08_MLPrediction.ipynb
│ ├── 09_MLComparison.ipynb
│ └── 10_confusionMatrix.ipynb
│
├── data/
│ ├── dataset_part_1.csv ← raw API data
│ ├── dataset_part_2.csv ← cleaned + Class
│ └── wiki_launches.csv ← scraped data
│
├── examResults/
│ ├── examGrade.png ← grading screenshot
│ └── AI_GradingFeedback.pdf ← AI evaluation report
│
└── presentation/
├── DS_Capstone_Coursera.pdf
└── DS_Capstone_Coursera.pptx
| Category | Tools |
|---|---|
| Language | Python 3 |
| Environment | Jupyter Notebooks |
| Data | Pandas, NumPy |
| Visualization | Matplotlib, Seaborn, Plotly |
| Maps & Dashboard | Folium, Plotly Dash |
| Machine Learning | scikit-learn (LogReg, SVM, DT, KNN) |
| Web Scraping | BeautifulSoup, Requests |
| Database | SQL / SQLite |
🤖 AI-assisted files
index.htmland thisREADME.mdwere generated with the help of artificial intelligence.🧠 Original work
All other content in this repository — including notebooks, analyses, models, and data — is entirely my own.
IBM Data Science Professional Certificate · Coursera · 2026